Estimation of continuous-time MIMO linear dynamic models from sampled data by hybrid parametrization

Publisher Website
Google Scholar
Abstract
The existing methods for estimating continuous-time MIMO linear dynamic models are mostly confined to the use of so-called indirect methods. However, such methods are possible only for those models that are parametrized in some special canonical forms, which usually do not embody the physical meaning of dynamic systems. Furthermore, the conversion from the discrete-time model to its corresponding continuous-time form is usually subject to limitations, and the results thus obtained are vague owing to the existence of stochastic noise. To circumvent the difficulties mentioned, a method that directly estimates the continuous-time MIMO linear dynamic model using hybrid parametrization is proposed. In this method, the continuous-time model can be parametrized according to the description of the physical system and then a discrete-time realization, which includes the parametrization of the stochastic noise, is formed as the basis for parameter estimation. This method is capable of estimating linear dynamic systems where both input disturbance and measurement noise, exist. It is also attempted to apply this method using the data from closed-loop operations for parameter estimations. Some numerical examples that enable comparison of the results thus obtained with those of some other methods are also included.

This publication has 28 references indexed in Scilit: